International variation in the management of severe COVID-19 patients

There is little evidence to support the optimal management of severe COVID-19 patients [1, 2]. To document whether there is a variation in practices, we performed an online survey (April 30–May 25, 2020) on behalf of ESICM.

In this online survey, a case vignette (https://​www.​surveymonkey.​com/​r/​F2FFC6S) was sent to ICU specialists who are members of ESICM. Questions investigated practices for a previously healthy 39-year-old patient presenting with severe hypoxemia from COVID-19 infection (Table 1). The 85 participating countries were grouped into 12 different regions [3]: continuous variables are described as median (interquartile range [IQR]) and are compared between groups using the non-parametric Wilcoxon rank-sum test. Categorical variables are described as frequency (percentages) and are compared between groups using Fisher’s exact test. Statistical analyses were performed with R statistical software, version 3.4.3 (available online at http://​www.​r-project.​org/​). A p value < 0.05 was considered significant.

As the patient had 88 (peripheral oxygen saturation) SpO₂ on 9 l/min of oxygen, direct ICU admission was reported in 56% (30–90%) of the respondents, with significant variation across regions (Fig. 1, P < 0.0001). Most intensivists not directly admitting patients to the ICU would admit them to an intermediate care unit managed by intensivists. However, the issue of bed availability was reported in South Europe (4.5%), South America (2.9%), Scandinavia (1.6%), Middle Europe (1.6%), and the UK (1.1%). Should the patient be not admitted to the ICU, a rapid response team would be involved in 29% of the cases, the ICU specialist would make the outreach her/himself in 24% of the cases, or an ICU nurse would be involved in 7% of the cases. In all other cases, ward or ED physicians would manage the patients. Direct ICU admission was significantly associated with baseline number of ICU beds (22 [12–40] vs. 18 [10–30] beds, P < 0.0001) and with the number of COVID-19 patients managed (30 [11–52] vs. 38 [20–70] patients, P = 0.001), as well as with management of patients in large cities (56.9% vs. 49.6%, P = 0.04).

Respondents were then asked about the first-line oxygenation strategy, which varied significantly across regions (Fig. 2, P < 0.0001). First-line high flow nasal cannula (HFNC) was used by 22.9% of the respondents (0% in Australia-New Zealand, 38% in Eastern Europe). Noninvasive ventilation was used by 25.5% of the respondents (5.4% in North America, 43.6% in the UK). Interestingly, 8% of the respondents were using first-line intubation (0% in Australia-New Zealand, 23% in Asia). Women less frequently initiated HFNC (32% vs. 42%, P = 0.02). The availability of an intermediate care unit influenced the use of HFNC or non-invasive ventilation (NIV) (32.8% vs. 21.7%, P = 0.03). Along this line, a higher number of ICU beds (24 (12–40) vs. 18 (10–30) beds, P = 0.0009) was associated with the use of HFNC and NIV. Interestingly, 37.5% were using prone positioning in awake non-ventilated patients. To assess whether HFNC or NIV should be continued, ICU specialists relied on SpO₂ (85.7%), respiratory rate (71.4%), followed by dyspnea (47.1%), and comfort (45.4%). Criteria for intubation included clinical signs of respiratory distress (94%), high oxygen flow to maintain a SpO₂ of 95% (33.5%), or low SpO₂ only (25.6%).

Following intubation, the patient had a partial pressure of oxygen/fraction of inspired oxygen (P/F) ratio of 84 mmHg. Although prone positioning (71.2%) and neuromuscular blockade (59.7%) were often used to optimize oxygenation, the practice varied significantly across countries. For instance, prone positioning was performed in 70–85% of the cases in Asia, India, Eastern Europe, Middle Europe, South America, South Europe, and the UK, whereas Africa, Australia-New Zealand, Middle East, North America, and Scandinavia were in the 50–70% range (Fig. 3, P < 0.0001). Respondent’s age was associated with the use of prone positioning (46 [39–54] vs. 44 [37–51] years, P = 0.007). Older respondent’s age (45 [37–52] vs. 47 [40–55] years, P = 0.0001), living in a large city (54.2% vs. 46.8%, P = 0.03), and a higher number of COVID-19 patients managed (35 [15–65] vs. 30 [12–55] patients, P = 0.02) were associated with the use of neuromuscular blockade.

Antibiotic prescribing was routine for all patients in 44.2% of the respondents and biomarker-guided in 36.5%, without significant variation across regions. Routine antibiotics were more frequently used by respondents working in university-affiliated hospitals (48.3% vs. 40.9%, P = 0.03) and those living in large cities (49.3% vs. 40.2%, P = 0.01). Biomarker-guided antibiotic therapy was less frequent in large cities (47.3% vs. 57.4%, P = 0.007). Regarding antiviral therapy, 48.9% reported not prescribing antivirals, 42.6% were giving hydroxychloroquine, 17% lopinavir-ritonavir, and 15% remdesivir. Figure 4 displays significant variation in antiviral prescriptions across regions (P < 0.0001). Physicians not prescribing antivirals were older (47 [40–54] vs. 44 [37–51] years, P < 0.0001), and more frequently men (55.4% vs. 39.9%, P < 0.0001). Conversely, those prescribing hydroxychloroquine were younger (43 [37–50] vs. 47 [40–54] years, P < 0.0001), and more frequently women (41.7% vs. 28.2%, P < 0.0001). There was significant variation in the use of interleukin-6 (IL-6)/IL-1 blockade or of corticosteroids across countries (P < 0.0001 for both tests). Other collected variables were not associated with the use of anti-inflammatory drugs.

This survey highlights important practice variations in the management of severe COVID-19 patients, including differences at regional and individual levels. This illustrates that neither IDSA nor Surviving Sepsis Guidelines did recommend any of these treatments, but instead encouraged inclusion of patients into trials [1, 4, 5]. Since the publication of these guidelines, no more evidence has been made available to ascertain that these specific COVID-19 therapies should be included in the standard of care. Learning from this heterogeneity will not only raise hypothesis on optimal patient’s management, but also serves as a tool to suggest personalized management for each clinical phenotype [6, 7].

This study has several limitations. First, the study suffers from a nonresponse bias of 80%. Second, even though only physicians have responded, we cannot ascertain that all of them had the clinical expertise and the experience of managing COVID-19 patients. Last, questions about specific treatments did not take into account the fact that the level of evidence has changed over time.

As no management guidelines have allowed to guide practices for the COVID-19 pandemic, heterogeneous behaviors are reported. Large outcome studies based on multinational registries are warranted.